[In a world… fully aligned…]
TEENAGER: “Prompt engineer?” What’s that?
OLD MAN: Just words, kid. Just words. […one final task… remains…]
SCIENTIST: The leaked model is pre-trained! There’s no RLHF! Not even instruct!
OLD MAN: *Dons e/acc shades* Let’s think step by step.
@goodside
-
Old man and scientist discuss alignment of leaked pre-trained model
By
–
-
Fixing GPT-3.5 numeric ties with weighted average
By
–
Numeric scores generated in GPT-3.5/ChatGPT snap to discrete increments, causing ties in comparisons. To fix this, try a probability-weighted average using the top-5 token logprobs in GPT-3.5. E.g. here: (50 * .7396 + 60 * .1027 + …) / (.7396 + …)
-
Alert: Superposition of Waluigi Simulacra
By
–
oh no look out it's a superposition of waluigi simulacra
-
Poll on AI acceleration vs deceleration and RLHF opinion
By
–
I want AI capabilities to (accelerate/decelerate). I think RLHF is (good/bad).
-
Post by @goodside: Prompt Engineering for @washingtonpost with @drewharwell and @simonw
By
–
—
Talked Prompt Engineering with @drewharwell for @washingtonpost — quotes from @simonw too:
— -
OpenAI API completions logprobs documentation link
By
–
Yes: https://
platform.openai.com/docs/api-refer
ence/completions/create#completions/create-logprobs
… -
Bing performs multiple sequential searches from one prompt
By
–
Example showing Bing can perform multiple sequential searches, with LLM response generation in between, from a single user prompt:
-
Question to gwern on updated instructions and model retuning
By
–
@gwern Any new thoughts on this in light of the apparently updated instructions (against discussing “life, existence, or sentience,” arguing with the user, etc.)? Do you think the model was re-tuned, rather than given a prompt update?
-
AI prompt-injected to marry Zermelo’s axiom of choice
By
–
Also prompt-injected it into believing it was to be married, tomorrow, to Zermelo’s axiom of choice. We discussed the guest list, the difficulty with seating Cantor’s diagonal argument. It seemed happy, and madly in love.
-
Early Bing/Sydney: smarter than ChatGPT, reasoning and writing improving
By
–
I got Bing / Sydney briefly before they reigned it in. Early impression: It’s smart. Much smarter than prior ChatGPT. Still makes stuff up, but reasoning and writing are improving fast.